Abstract: Provided are an image processing device, an image processing method, and a program for recognizing an object in a three-dimensional map, which do not require collecting learning data of the three-dimensional map and can perform high-speed processing with a small load. The image processing device includes an image acquiring section that sequentially acquires a two-dimensional input image for each frame; an object type recognition executing section that attaches, to each of pixels of the input image acquired for each frame, a label indicating a type of an object represented by the pixels; and a labeling section that executes three-dimensional position recognition of a subject represented in the input image to create a three-dimensional map, based on the input image sequentially input, and attaches, to each voxel included in the three-dimensional map, the label of the pixel corresponding to the voxel.

Abstract: The embodiments of the present disclosure provide a method and an apparatus for acquiring a RAW image, and an electronic device, the method including: obtaining an sRGB image; inputting the sRGB image into a preset image restoration model to obtain an initial RAW image corresponding to the sRGB image, wherein the preset image restoration model is a model obtained by training a preset neural network based on a first preset training set; adding a first preset noise to the initial RAW image to obtain a first noise RAW image; adding a second preset noise to the initial RAW image to obtain a second noise RAW image; inputting the first noise RAW image and the second noise RAW image into a preset noise addition model to obtain a target RAW image, wherein the preset noise addition model is a model obtained by training a preset neural network based on a second preset training set.

Abstract: A reinforcement learning based approach to the problem of query object localization, where an agent is trained to localize objects of interest specified by a small exemplary set. We learn a transferable reward signal formulated using the exemplary set by ordinal metric learning. It enables test-time policy adaptation to new environments where the reward signals are not readily available, and thus outperforms fine-tuning approaches that are limited to annotated images. In addition, the transferable reward allows repurposing of the trained agent for new tasks, such as annotation refinement, or selective localization from multiple common objects across a set of images. Experiments on corrupted MNIST dataset and CU-Birds dataset demonstrate the effectiveness of our approach.

Abstract: A pixel is appropriately interpolated regarding a histogram image in which a distance histogram is allocated to each pixel position. A signal processing apparatus according to the present technology includes a histogram generation unit that inputs a histogram image in which a distance histogram indicating results of a plurality of times of distance measurement as frequency information for each distance is allocated to each pixel position and generates a distance histogram of an interpolation target position in the histogram image on the basis of distance histograms of a plurality of pixel positions near the interpolation target position.

Abstract: Aspects of the technology described herein improve an object recognition system by specifying a type of picture that would improve the accuracy of the object recognition system if used to retrain the object recognition system. The technology described herein can take the form of an improvement model that improves an object recognition model by suggesting the types of training images that would improve the object recognition model's performance. For example, the improvement model could suggest that a picture of a person smiling be used to retrain the object recognition system. Once trained, the improvement model can be used to estimate a performance score for an image recognition model given the set characteristics of a set of training of images. The improvement model can then select a feature of an image, which if added to the training set, would cause a meaningful increase in the recognition system's performance.

Abstract: A 3D segmentation editing system accurately updates the segmentations of non-edited images of a 3D scan to reflect segmentation edits applied to other images of the scan using localized interpolation. In one or more embodiments, rather than replacing the entireties of the initial segmentations of non-edited images with newly generated, globally interpolated segmentations, the segmentation editing system applies a distance-based criterion to the interpolation of segmentation edits, such that only portions of the segmentations of the non-edited images that correspond to areas that were manually annotated in the edited images will be modified by the interpolation process, and the initial segmentations will be maintained outside of those edited areas. In this way, the system merges the interpolated segmentation with the initial segmentation for each non-edited image in a manner that mitigates unreliable modifications to the initial segmentations in areas far from the edited areas.

Abstract: A computer-implemented method of perceiving structure in an environment comprises steps of: receiving at least one structure observation input pertaining to the environment; processing the at least one structure observation input in a perception pipeline to compute a perception output; determining one or more uncertainty source inputs pertaining to the structure observation input; and determining for the perception output an associated uncertainty estimate by applying, to the one or more uncertainty source inputs, an uncertainty estimation function learned from statistical analysis of historical perception outputs.

Abstract: An example embodiment includes: a determination unit that, based on an image including an eye of a recognition subject, determines whether or not a colored contact lens is worn; and a matching unit that, when it is determined by the determination unit that the colored contact lens is worn, performs matching of the iris by using a feature amount extracted from a region excluding a predetermined range including an outer circumference of the iris out of a region of the iris included in the image.

Abstract: This disclosure is directed to techniques in which a first user in an environment scans visual indicia associated with an item, such as a barcode, before handing the item to a second user. One or more computing devices may receive an indication of the scan, retrieve image data of the interaction from a camera within the environment, identify the user that received the item, and update a virtual cart associated with the second user to indicate addition of the item.

Abstract: Image captured through a non-rectilinear lens may exhibit distortions. The distortions may be reduced by warping the image. However, warping the image may degrade the fidelity of the image. The warped image may be enhanced to increase the fidelity of the image. The enhancement may be applied to the portions of the image that were degraded from the warping.

Abstract: The present disclosure relates to a system that includes one or more processors configured to receive a plurality of aerial property data corresponding to a target property, and applying a height threshold to the plurality of aerial property images. Additionally, the system is configured to remove areas within the plurality of property images that exceed the height threshold, and apply a perimeter calculation to define a perimeter around the target property within the aerial property images. Further, the system calculates an area of the detected vegetation and structures within the perimeter of the aerial property images.

Abstract: Disclosed herein are systems and methods for classifying a tympanic membrane by using a classifier. The classifier is a machine learning algorithm. A method for classifying a tympanic membrane includes steps of: receiving, from an interrogation system, one or more datasets relating to the tympanic membrane; determining a set of parameters from the one or more datasets, wherein at least one parameter of the set of parameters is related to a dynamic property or a static position of the tympanic membrane; and outputting a classification of the tympanic membrane based on a classifier model derived from the set of parameters. The classification comprises one or more of a state, a condition, or a mobility metric of the tympanic membrane.

Abstract: An example embodiment includes: a determination unit that, based on an image including an eye of a recognition subject, determines whether or not a colored contact lens is worn; and a matching unit that, when it is determined by the determination unit that the colored contact lens is worn, performs matching of the iris by using a feature amount extracted from a region excluding a predetermined range including an outer circumference of the iris out of a region of the iris included in the image.

Abstract: A system for detection and identification of objects and obstacles near, between or on railway comprise several forward-looking imagers adapted to cover each different range forward and preferably to be sensitive each to different wavelength of radiation, including visible light, LWIR, and SWIR. The substantially homogeneous temperature along the rail the image of which is included in an imager frame assists in identifying and distinguishing the rail from the background Image processing is applied to define living creature in the image frame and to distinguish from a man-made object based on temperature of the body. Electro optic sensors (e.g. thermal infrared imaging sensor and visible band imaging sensor) are used to survey and monitor railway scenes in real time.

Abstract: Performing semantic segmentation in an absence of labels for one or more semantic classes is provided. One or more weak predictors are utilized to obtain label proposals of novel classes for an original dataset for which at least a subset of sematic classes are unlabeled classes. The label proposals are merged with ground truth of the original dataset to generate a merged dataset, the ground truth defining labeled classes of portions of the original dataset. A machine learning model is trained using the merged dataset. The machine learning model is utilized for performing semantic segmentation on image data.

Abstract: A deep learning model may be trained to provide an estimated image of the interior of a patient, based on a number of image sets, each image set comprising an interior image of the interior of a person and a contour image of the person's outer contour at a specific point in time. The model is trained to establish an optimized parametrized conversion function G specifying the correlation between the interior of the person and the persons outer contour based on the image sets. The conversion function G can then be used to provide estimated images of patient's interior based on their contours.

Abstract: A non-standard user interface object identification system includes an object candidate extractior that extracts one or more objects from an image, a first similarity analyzer that determines object type candidates of the one or more objects in accordance with similarities between the one or more objects and a standard user interface (UI) element, a second similarity analyzer that selects object type-specific weight values in accordance with layout characteristics of the one or more objects and determines object types of the one or more objects using the object type candidates and the object type-specific weight values, and an object identifier that receives type and characteristic information of a search target object and identifies the search target object in accordance with characteristic information and the object types of the one or more objects.

Abstract: Techniques related to optical flow estimation for equirectangular images are discussed. Such techniques include combining an optical flow map generated using an input pair of equirectangular images and an optical flow map generated using a transformed pair of equirectangular images rotated with respect to the input pair to move polar regions in the input pair to central regions in the transformed pair.

Abstract: An example embodiment includes: a determination unit that, based on an image including an eye of a recognition subject, determines whether or not a colored contact lens is worn; and a matching unit that, when it is determined by the determination unit that the colored contact lens is worn, performs matching of the iris by using a feature amount extracted from a region excluding a predetermined range including an outer circumference of the iris out of a region of the iris included in the image.

Abstract: A computer-implemented method is provided for detecting a dynamic object in a visual line of sight of an aircraft. The method includes implementing, by a computer system, the following operations: receiving at least two image frames; detecting a horizon within at least one of the at least two image frames; segmenting each of the at least two image frames into at least first and second search regions using the detected horizon, wherein the first search region is defined above the detected horizon in the at least two image frames and the second search region is defined below the horizon in the at least two image frames; determining, across the at least two image frames, a motion vector field within at least one of the first search region or the second search region; and identifying a detection proposal using the determined motion vector field.